DEPTH EXTENT OF INNER-CORE SEISMIC ANISOTROPY AND IMPLICATIONS FOR GEOMAGNETISM

To constrain the elastic structure of the Earth's inner core, we have picked the differential times of 879 core-penetrating body waves from vertical-component, short-period seismograms recorded by global and re gional seismic networks. Using a cross-correlation technique, we measu re the difference in arrival times of P'(BC)-P'(DF) and P'(AB)-P'(DF) where the P'(DF) (PKIKP) phase penetrates the inner core, while both t he P'(BC) (PKP-BC) and P'(AB) (PKP-AB) phases bottom in the outer core , P'(BC)-P'(DF) times for paths that are nearly parallel to the Earth' s spin axis are consistently 2-4 s larger than predicted using the Pre liminary Reference Earth Model (PREM), while rays in other directions have a mean and standard deviation of 0.2 +/- 0.4 s. P'(AB)-P'(DF) tim es, which correspond to P'(DF) rays turning deeper in the inner core, are 3-6 s for rays nearly parallel to the spin axis and 0.3 +/- 0.9 s for rays not near the spin axis. These observations lead to the robust conclusion that the inner core is strongly anisotropic. The level of anisotropy in our model is about 3% at a radius of 1000 km (depth of 2 00 km) and increases to about 4% at a radius of about 700 km. Below th is radius, our resolution is poor, but the anisotropy appears to weake n. Resolution is also weak in the outer 200 km of the inner core, but the anisotropy appears to diminish in this region as well, A simple mo del of hexagonally symmetric anisotropy aligned with the spin axis exp lains 74% of the variance. The symmetry direction which fits the data the best and explains 79% of the variance is at 80 degrees N, 120 degr ees E. The locus of directions which explain 70-79% of the variance in cludes only 4% of the possible range of directions, and includes the s pin axis direction. The observed anisotropy is most likely due to pref erred alignment of elastically anisotropic crystals, We propose severa l new alignment mechanisms and all viable mechanisms seem to be associ ated with a strong toroidal magnetic field. An outstanding problem tha t requires further investigation is that first-principles calculations of seismic anisotropy of hexagonal close packed (hcp) iron suggest th at anisotropy of order 3% is predicted. Thus, 100% alignment could go a long way towards explaining our observations, but seems highly unlik ely.